What Large-Scale 5G Trials Are Teaching Us About 6G

Discussions about 6G often concentrate on the technologies that will differentiate the next generation of mobile networks: Integrated Sensing and Communication (ISAC), AI-native networks, new spectrum bands, semantic communications, distributed intelligence and new device types.

However, some of the most useful lessons about 6G may actually be coming from large-scale 5G deployments taking place today.

The 6G Infrastructure Association (6G-IA) Trials Working Group published a white paper in June 2026 titled “Large-Scale Trials in SNS JU Projects – Technical Insights”. Rather than focusing on a single research project, the paper brings together findings from 27 large-scale trials and pilots across 10 European countries, six vertical domains, nine SNS projects and more than 21 testbeds.

The trials cover a broad range of applications including public safety, agriculture and forestry, automotive and transportation, healthcare and education, immersive media and XR, Integrated Sensing and Communication, smart cities and industrial applications. Public Protection and Disaster Relief (PPDR) represents the largest individual category, accounting for eight of the 27 trials.

What makes the white paper particularly interesting is the final part, where the authors look across all the individual projects and identify common technologies, recurring problems and lessons for future 6G trials.

One of the clearest findings is that 5G Standalone has effectively become the foundation for advanced vertical applications.

According to the analysis, 5G SA appeared in more than 25 of the 27 trials. Edge computing or MEC appeared in 21, end-to-end network slicing in 19, AI/ML at the edge in 16 and Kubernetes-based orchestration in 13. In comparison, ISAC or JCAS, which we normally associate more directly with future 6G capabilities, appeared in only two trials.

In other words, before many of the more futuristic 6G capabilities can become useful, there is already a fairly consistent underlying platform emerging:

5G SA + Network Slicing + Edge Computing + Cloud-native orchestration + AI/ML

The report goes as far as describing 5G SA and end-to-end slicing as essential rather than optional for differentiated service quality. Trials using NSA or unmanaged shared resources experienced greater latency variation and reduced reliability under congestion.

There are some useful real-world examples.

During the FIDAL PPDR trial in Málaga, a dedicated public-safety slice running on Telefónica's commercial 5G SA network maintained Mission Critical Push-to-Talk access performance during congested stadium conditions, while ordinary users on the default slice experienced degraded throughput.

At the La Mayora experimental farm, three slices were used simultaneously: an eMBB-type slice for 4K UAV video, a low-latency slice for drone control and an mMTC-type slice for agricultural sensors. The trial demonstrated that these services could share the same physical infrastructure while maintaining resource isolation.

The trials also reinforce something that has been discussed since the early days of 5G: latency is not simply a radio-network problem.

Every trial in the study requiring latency below 50 milliseconds reached essentially the same architectural conclusion: processing needs to move closer to the user.

One connected-vehicle trial at IDIADA near Barcelona measured around 24 ms end-to-end latency using 5G and edge computing, compared with approximately 106 ms using 4G and a centralised cloud implementation. The report characterises this as roughly a four-to-five-times improvement.

A public-safety AR trial similarly achieved around 30–40 ms end-to-end latency while processing video using edge-based AI. Moving that processing further into a central cloud would have pushed latency beyond what was considered acceptable for the AR application.

This is important for 6G because ultra-low latency will increasingly depend on where compute, storage and AI functions are located, not simply on improvements to the air interface.

AI is another interesting finding.

The statistical analysis identified AI/ML at the edge in 16 trials, and the report argues that AI has already moved from being an optional network enhancement to becoming part of the application itself. Examples include fire detection, maritime surveillance, vehicle anomaly detection and agricultural image classification.

This distinction matters.

Much of the discussion around AI-native 6G concentrates on using AI to operate or optimise the network. These trials show the other side of the equation: future networks will also increasingly carry applications where real-time AI inference is an integral part of the service.

The network, compute platform and AI workload therefore have to be considered together.

Another lesson is that uplink capacity remains a significant challenge.

Trials involving drones, body cameras and multiple simultaneous video streams repeatedly identified the uplink as a bottleneck. One FIDAL digital-twin trial achieved an aggregate uplink of around 120 Mbps while supporting two drones and four surveillance cameras simultaneously. Dedicated slicing and resource allocation were needed in other scenarios to prevent high-bandwidth UAV video from competing with sensor traffic.

This could become even more significant towards 6G.

Many future applications are likely to involve machines generating data rather than simply consuming it: drones, vehicles, robots, cameras, digital twins and increasingly sophisticated sensors. The traditional assumption that mobile networks primarily need very large downlink capacity becomes less useful in this environment.

Interestingly, some of the biggest limitations encountered during the trials had little to do with the network itself.

The white paper identifies the lack of native 5G SA devices for drones, XR headsets and industrial equipment as the most frequently reported technical limitation. Researchers often had to use USB adapters, Ethernet converters, Wi-Fi tethering or external 5G routers. These workarounds introduced additional latency and jitter and sometimes made accurate measurement more difficult.

That is an important reminder for 6G development. A standards-compliant network does not create an ecosystem by itself. Devices, modules, chipsets, applications and industrial interfaces all have to mature alongside the infrastructure.

The operational findings may be even more valuable than some of the technical results.

Projects involving emergency services, healthcare workers and industrial users found that real users need to participate in designing the trials rather than simply being brought in at the end to evaluate them. Early involvement changed scenarios, interfaces and operational procedures in ways researchers had not anticipated.

The paper also concludes that in areas such as PPDR and healthcare, the main obstacle was frequently technology adoption rather than network performance. Reliability, simple interfaces, training and integration into existing operational processes were often more important to users than having additional features.

This is something the telecoms industry occasionally forgets when discussing future network generations.

A technically impressive 6G capability has limited value if the intended users cannot easily incorporate it into their existing workflows.

Large-scale trials also require considerable preparation. According to the report, permits, safety procedures, device procurement, SIM provisioning and agreements with operators often require four to six months or more. In operational environments such as ports, hospitals, stadiums and transport facilities, logistics rather than network integration can become the critical path.

Measurement itself is another recurring challenge.

Synchronised timestamps across devices, networks, applications and edge platforms were found to be essential for meaningful end-to-end KPI measurements. Interestingly, several projects independently converged on combinations of Prometheus, InfluxDB and Grafana for observability and time-series analysis. More than eight projects adopted similar monitoring approaches without central coordination.

The report therefore recommends that future 6G trial programmes standardise more of the underlying reference architecture and measurement environment rather than repeatedly rebuilding them for every project.

The proposed starting point is revealing: 5G SA, dedicated network slicing, distributed edge UPF, Kubernetes, AI/ML inference, API-based QoS control and common observability tools. Future projects could then concentrate their research effort on the genuinely new 6G layers built above this platform.

The report also highlights several challenges that remain unresolved.

Moving towards fully disaggregated B5G and 6G architectures can introduce additional latency through the increasing number of interfaces and distributed components. Industrial applications continue to suffer from limited interoperability between tools such as BIM platforms, robotics systems and Mixed Reality devices. Some Key Value Indicators, such as long-term societal benefits or reductions in casualties, are also impossible to validate within the timescale of a conventional technology trial.

Finally, the white paper recommends that future trials begin systematically combining the technologies that more clearly point towards 6G.

These include ISAC, AI-native control loops, intent-based networking, semantic communications and energy-harvesting IoT. The existing JCAS/ISAC work and intent-based orchestration experiments are described as early 6G proof points that can build on the now much more mature 5G platform underneath them.

Perhaps the biggest message from the report is therefore that the transition from 5G to 6G is unlikely to be a clean break.

The large-scale trials suggest that many of the foundations of future 6G networks are already becoming visible through advanced 5G deployments. 5G Standalone, slicing, edge computing, cloud-native orchestration, APIs, observability and AI are gradually forming a common platform.

6G will add new capabilities on top of this, particularly sensing, deeper integration of AI, new forms of automation and potentially entirely new ways of communicating and interacting with the physical world.

But before those capabilities can be deployed at scale, the industry still needs to solve some surprisingly practical problems involving uplink capacity, device availability, interoperability, measurement, operational processes and user adoption.

In that sense, the most useful preparation for 6G may not always be another laboratory demonstration of a futuristic radio technology.

It may be making the advanced 5G platform underneath it robust enough to support what comes next.

Related Posts: 

Posted by on

Comments

Post a Comment